Quite an unusual setup, as you'll need a gasket under each nut.
Anyway I think that you should be able to use App.2 of Div.1 without changes, some lever arms will be negative and you should conserve the sign in the equations.
If I understand the issue correctly, the proposed flange is similar to that shown in VIII-1 App 2 Fig 2-13 with the bolt hole and gasket surface reversed. I haven't pondered the implication much, but seems that the internal moment caused by the bolt-gasket couple is behaving the same way as it would for a normal flange. Thus, I'd start with a "normal" flange calc. The only thing substantially different is the flange is relatively smaller and, of course, needs to have internal access on both sides for bolt-up and will interfere with the fluid flow like an RO would.
Sorry for the unclearness...but it isn't a 2.13 type.
You could see it as a regular loose type flange (type 2.4) but with the gasket outside the bolts & the flange is positioned in the middle of a pipe.
The process fluid is around the flange and pipe.
flaka, can you post a figure of your flange? (see FAQ559-1177). I can't see how you can make the flange pair tight, as the bolt holes can be tapped on one flange, but cannot be on the mating one.
This flange pair could be considered to be subject to external pressure, and thus calculated per 2-11, but, as the gasket is outside the bolt circle, it behaves like a flange subject to internal pressure, with all three loads H[sub]D[/sub],H[sub]G[/sub],H[sub]T[/sub] acting in the same direction, thus adding up in the moment equation.
In my opinion, changing my position with respect to my previous post, you should calculate flange moments per 2-6 using h[sub]G[/sub]=(G-C)/2 and M[sub]0[/sub]=W(G-C)/2